22th North America-Europe Data Exchange Meeting Reading, UK, Dec. 09-11, 2009

22 th NAEDEX Meeting Reading Alexander Cress

Deutscher Wetterdienst

22th North America-Europe Data Exchange Meeting Reading, UK, Dec. 09-11, 2009

Deutscher Wetterdienst (DWD) status report

Alexander Cress

With contributions from Rheinhold Hess, Andreas Rhodin, Marco Schwaerz, Klaus Stephan



DWD’s new building 2008

New central computer facilities:• Computer area 1100 m2 • Uninterruptible power supply up to 3000 kW



• 102.4 GFlops / CPU

• 1.6 TFlops / Node

• 512 GB Main Memory per node

• 16 CPUs per node

• Frequenzy 3.2 GHz

• Ultra-High bandwidth shared memory

subsystem (256 GB/s per CPU)

• Internode crossbar switch (IXS),

128 GB/s bidirectional per node

• DWD 14 nodes per cluster, 2 clusters, 1 test node

New computer system at DWD Vector Computer NEC SX-9

System specification



Global Model GME

• Operational NWP Model of DWD

• gridpoint model, hexagonal triangular grid

• 40 km mesh size, 36870 grid points/layer

• 40 layers (hyprid, sigma/pressure)

• prognostic variables: ps, u, v, T, qv, qc, qi, o3

• 3DVAR (PSAS) system

• incremental digital filter initialization (P.Lynch)

• At 00 UTC and 12 UTC: forecasts for 174 hours

• At 18 UTC: forecasts for 48 hours



3-D Variational Data assimilation (PSAS)

Replacement of the Optimum Interpolation (OI) scheme

Operational since Sep. 2008 and operational on NEC SX9 since Sep. 2009

Minimization in observation space

Allows assimilation of observations with highly nonlinear dependence on the background variables (e.g. remote sensing)

Wavelet representation of the B-Matrix• Separable 1D+2D approach• vertical: NMC derived covariances (64 levels)• horizontal: Wavelet representation (512x256 Gaussian grid)

Goals:• Flexible representation of B-Matrix• 3D non-separable representation• NMC or ensemble derived covariances



Observational Use in 3DVAR

Aircraft Synop Buoy

Radiosonde Pilot AMSU A

AMV Scatterometer


Alexander Cress

Anomaly correlation coefficient of geopotential height in 500 hPa2007051400 – 2007053100 (18 forecasts)


Alexander Cress

Use of AMSU-A auf NOAA-15, 16, 18, AQUA und METOP-A (seit 21.11.07)


Using Metop ATOVS radiances: Clear positive impact for all areas (00 UTC)



Use of radio occulation measurements

A electromagnetic signal, transmitted by a GPS satellite, is delayed due to the presence of free electrons in the ionosphere an the refractivity of the atmosphere

The refractivity is directly linked to horizontal and vertical variation of temperature, pressure and water vapour

The refraction of the GPS signal corresponds to a shift in its phase, recorded at the receiving LEO satellite. Additionally, the signalpath undergoes a bending in the atmosphere, resulting in a bending angle

Relative geometry of the GPS and LEO and the Earth changes during the occultation event, the signal path intersects the atmosphere vertically, thus providing a vertical profile of bending angle.

Observation error can be derived from the signal-noise-ratio of the amplitude of the signal

Observed quantity to be assimilated can be bending angle (DWD) or refractivity (Met Office) profiles

Benefits: High vertical resolution, independence of cloud conditions, lack of fundamental biases, uniform global coverage

Observation coverage





Read SCATT (25 km BUFR data)Data from data base or file or ECMWF MARS System

Pre SCATT- read in original Bufr- data selection within analyses time window- eliminates overlapping orbits- computes KNMI rain flag; data flaging- computes direction check; data flagging- computes bias correction- data selection based on quality control - thinning- output options: Bufr; netCDF, ascii

OI - Analyses- (u/v) most likely wind- coded as Pseudo-Buoy (Bufr)

3DVAR - Analyses- (u/v) more than one wind- coded as Bufr/netCDF

Offline Monitoring- colloc. with GME -FG/Ana- analyses of data quality

Use of scatterometer data at DWD

• 10 m wind vectors (most likely wind)

• QuikScat and ASCAT

• Global and regional

• Use of multiple wind solutions (planned)

• Experiments with OI and 3DVAR



Scatterometer Data Coverage2008022500 +/- 1.5 H

ASCAT (red) QuikScat (blue)


Deutscher WetterdienstScatterometer data in the ParallelroutineTropical Storm 02B 2009052412 sea level pressure [hPa]

Routine Parallelroutine

Roup - Rou ECMWF


Deutscher WetterdienstScatterometer data in the Parallelroutine

Tropical Storm 02B 2009052412 vv=12h sea level pressure [hPa] / max windspeed [m/s]

Routine Parallelroutine



CrtlCrtl plus SCAT

CrtlCrtl plus SCAT

Anomaly correlation coefficient for sea level pressure

Northern Hemisphere Southern Hemisphere

EuropeVV=72h

CrtlCrtl plus SCAT



Use of the polar AVHRR wind vectors in the global analyses system of DWD

• Polar regions with small observation density

• Polar regions show still large observation errors

• Polar lows have influence on weather regimes

in Europe and North America

• Polar AMV winds from Terra and Aqua only

Experimental

• No operational satellite programm planned for

• Derivation of AMV wind vectors in polar regions

• Deriation of wind vectors from polar orbiting

• NOAA satellites (15/16/17/18) and Metop

• Only infrared winds

• Height assigment more problematic than for Modis winds



Metop NOAA 17

NOAA 18 Modis/Terra

AVHRR OBS – FG Statistik 2008081400 – 2008083121

All Data mit QI > 65 700 hPa – 400 hPa



Anomaly correlation coefficient of the 500 hPa geopotential height2008100200 – 2008102300 00 UTC 22 cases

Control (red) Exp. with AVHRR Winden (blau)



Time series of anomaly correlation coefficients 96-h forecast of the 500 hPa geopotential height field

2008100200 – 2008102300 00 UTC

NH TR

SH EU

Routine Exp. Mit AVHRR



Direct Broadcast MODIS Winds

• MODIS polar winds are not available in time to be used in assimilation of main run. Only available in assimilation run

• Direct broadcasting winds can be received much earlier ~ 100 minutes or more • Winds from a variety of stations

Tromso - Terra Modis Sodankyla - Terra Modis Fairbanks - Terra Modis McMurdo, Antartica - Terra/Aqua Modis

• Provide only partial coverage and only Terra can be received in the NH

• At DWD, no MODIS winds could be used in the main runs. Using DB winds, some polar winds can be used also in the main run. Additionally, more polar winds can be used in the assimilation

Motivation



Alexander Cress

Data coverage

00 UTC

12 UTC



StoreTime

Nu

mb

er o

f O

bse

rvat

ion

s



OBS –FG StatisticsTerra QI > 65

20090111 - 20090119IR 700 – 400 hPa IR 400 – 0 hPa

WV 700 – 400 hPa WV 400 – 0 hPa

Mean: -0.05

Mean: -0.26

Mean: -0.13

Mean: 0.38



Alexander Cress



Alexander Cress


ww

Lokal-Model COSMO-EU (LME) und COSMO-DE (LMK)

COSMO-EU (regional model): non-hydrostatic, rotated lat-lon grid, mesh-size: 7kmterrain-following hyprid coordinate with 40 layers up to 20 hPaforecast range: 78 h every 6 hoursprognostic cloud ice, prognostic rain schemesboundary values from GME

Analysis: continuous nudging schemeobservations: radiosonde, pilots, wind profiler,

aircraft, synops, buoys, shipscut-off: 2h30minvariational soil moisture analysis

COSMO-DE (lokal model): similar to COSMO-EUforecast range 18 h every 3 hmesh-size: 2.8 km, explicit convectionlatent heat nudging of radar reflectivitiesboundary values of COSMO-EU

GME

COSMO-EU

COSMO-DE



Assimilation of Scatterometer Wind in COSMO-EU

Heinz-Werner Bitzer (MetBw), Alexander Cress, Christoph Schraff (DWD)

nudging of scatterometer wind data technically implemented, taking into account all quality control / bias correction steps developed for use in GME

idealised case studies: model rejects largest part of 10-m wind info unless mass field is explicitly balanced

derive surface pressure analysis correction in geostrophic balance with 10-m wind analysis increments (implies need to solve Poisson equation):

implemented, model now accepts data

first real case study computed QSCAT 19 June 2007, 6 – 9 UTC

48N

50N

15 W

Opr (no QSCAT) – Exp (QSCAT)PMSL 19 June 2007, 9 UTC hPa


Deutscher Wetterdienstwith ASCAT / QuickScat

pmsl (model – obs)

too low

too strong

gradient

COSMO-EU

9-h forecasts,

valid for

6 March 2008,

9 UTC

No scatt



COSMO-EU ana with ASCAT/QuickScatCOSMO-EU ana , no scatt

ECMWF analysis 29 Feb 08ASCAT 28 Feb 08, 21 UTC ± 1.5h

984 hPamax. 30 kn

~15 m/s

10-m wind [m/s]

975 hPa



Compariosn of surface weather elements between COSMO-EU Routine and COSMO-EU experiment including Scatterometer data

27/02/2008 – 09/03/2008 00 UTC

Routine

Exp. with Scatt



Assimilation of IASI Measurements into the COSMO-EUMichael Schwärz

EUMETSAT Fellow

• Infrared atmospheric sounding interferometer onboard Metop

• IFOV: 3.33o (48 km nadir)

• Swath: +/- 1026 km

• 8461 channels ⇨ 300 channels selected by IC

• Use of RTTOV 9 within 1DVAR

• Bias correction (Harris and Kelly 2001)

• Cloud detection a) IASI level 2 cloud flags b) after McNelly and Watts (2003)

• Use of temperature and humidity profiles in COSMO EU



Experiment design

• 215 temperature channels from 15 µm band• 6.25 µm wv band• IASI level 2 flags for cloud detection• COSMO-EU + IASI 1DVAR profiles

Results• Data processing and nudging works • Positive results in upper air verification• Stronger for RMS than Bias• Heighest for geopotential height in the upper troposphere and humidity in the middle troposphere

OutlookBetter channel selectionThinning in COSMO – EUUse of cloud detection by McNelly and Watts



COSMO-DE 2.8 km solution with 50 vertical layers

Runge Kutta time integration

Explicit deep convection/ param. shallow conv.

Rain, snow, graupel

Started every 3 hours (30 min cut off)

Forecast range 21 h

Nudging of conventional data

Latent Heat Nudging of radar derived rain rates (16 German stations, will be extended to about 30 stations, soon)

Domain of COSMO-DE andGerman Radar Network

COSMO-DE

To provide the nowcasters a appreciate guidance on severe weather events related to deep moist convection (super- and multi-cell thunderstorms) or to small scale orography



Skill scores for a convective period (14 days) for 0.1 mm/h

about midnight (00 UTC) about noon (12 UTC)

Comparison of forecast starting with and without LHN

LHNnoLHN

LHNnoLHN

LHNnoLHN

LHNnoLHN

Forecast starting from LHN analysis are betterLHN via noLHN

ET

SE

TS

FB

IE

vent

s



Room for improvement: data base

Up to now used withtin LHN:

All 16 German stations

Shortly to be extended with

3 Dutch stations

2 Belgian stations

10 France stations

3 Swiss stations

Quality Control Clutter filter Cross error detection Bright band correctio blacklisting (comparison to satellite picture)



Impact of extended data: Hourly Precipitation on 25.05.2009 12 UTC +5h

New Radar CompositeForecast with orig. data Forecast with new data

Forecast is improved by extended data base

Extended Data Base



Summary

• 3DVAR outperforms OI and is operational since mid Sep. 2008

• Positive impact of ATOVS radiances from Metop and NOAA 19

• Use of radio occulation in 3DVAR showed some promising results • Positive impact of scatterometer data in global and regional forecasts

• Substantial benefit of AVHRR and DB MODIS polar winds

• First results of using IASI data in COSMO-EU promising

• LHN of radar derived rain rates is beneficial for the forecast, esp.for nowcasting purposes



Plans for the next years• Continuation of experiments using Radio Occultation data including Metop

• Use of AMSU B & MHS data (first experiments running)

• Use of AIRS & IASI data in GME (already started)

• Prepare for ADM mission

• Use of VAD and radar radial winds in COSMO Models

• Use of GPS humidity data (COPS reanalysis)

• Development of a new non-hydrostatic Global Model with local zooming (ICON)

• Development of an Ensemble Kalman Filter (LETKF) for the new model ststem



Thank you for your

attention

Documents

22th North America-Europe Data Exchange Meeting Reading, UK, Dec. 09-11, 2009